Evaluation Conclusion

This chapter served to answer the research sub-question: “​How can the validity of the classification be verified?​”

The program was tested using the Unit Test method where single components of the code are tested. After that experts in the field of data analysis and breathing were approached to comment on the validity of the methodology and approach to classify breathing data. Lastly, the program as a whole was tested on participants with prior diaphragmatic breathing experience.

The Unit Tests were all conducted successfully, and showed that the program was able to handle errors in the data and user input, as well as various scenarios for calculations, such as the rolling over of an hour or day while calculating the difference in time. The experts that were approached validated the system in terms of methodology, the way the data was processed and the approach to classify breathing. They also made some suggestions for future work. The results of the DBM detection test

8. Conclusion

This chapter marks the end of the report, and answers all the research sub-questions listed in Section 1.3 in brief, followed by an answer to the main research question.

8.1 Conclusion

This section will begin by answering, in short, the sub-research questions formulated to help answer the main research question. The first four sub-research questions address the topics of the health benefits of diaphragmatic breathing, the use of RIP as a measurement for breathing, the classification of respiratory data and habit formation, for background information. Using this information, and the information gathered by answering the remaining sub-research questions regarding the kinds of features users like to be provided and testing the validity of the developed program, the main research question is answered.

“What are the physiological health benefits of diaphragmatic breathing?”

By doing literature research, it was found that diaphragmatic breathing has many applications as interventions and treatments to conditions such as oxidative stress, anxiety and hypertension.

Diaphragmatic breathing needs to be practiced regularly, for periods of longer than 4 weeks to observe sustained and consistent results. Motivation plays a role in making users stick to practice routines.

“Why is RIP used for recording breathing data?”

From literature, it was found that RIP has been validated as method to measure respiration, and that the position of the bands varies between studies, however is generally at the level of the axilla and navel. It was concluded that it is not an all purpose breathing measurement technique and suffers during activities involving large amounts of physical movement and activity. Slippage of the thoracic band was found to be a major source of error as can be confirmed from the results of the Program Verification test.

“How can respiratory data be classified?”

The types of classifiers used for breathing pattern recognition vary between studies. The classifiers used depended on the complexity of the data being recorded and the goal of the classification. The features extracted from the data were loosely dependent on the goal of the studies. Some studied focused more on time domain components while others on those from the frequency domain. Each classifier excelled at performing certain types of classification and one cannot be used to do everything.

“How is a habit formed?”

Habits are formed when an action is repeatedly done as a response to a certain cue or context. The amount of time a habit takes to form varies significantly between individuals and is dependent on their motivation to form the habit and the independence they are given to form said habit. Habits can weaken over time if not performed frequently and is a point to reflect upon when weaning users off the

Airleviate, to perform diaphragmatic breathing on their own.

“_​What kinds of breathing features would users like to be presented for the purpose of feedback?​” From the brainstorms, it was found that users are interested in statistics like the a quality factor to describe the performance during a session, the longest DBM and the average DBM duration. Similar features were thought up during the in individual brainstorm, showing that the author and potential users have similar ideas in mind for what the breathing feedback should consist of. The interviews showed that the client had an interest in moments of rest (where the user is relatively stationary) and moments of periodic breathing (where the user’s breathing frequency does not vary too much with time). The complete list of features can be found in the Preliminary Requirements in Section 4.8.

“How can the validity of the program be verified?”

The program was tested using the Unit Test method where single components of the code are tested. The Unit Tests were all conducted successfully, and showed that the program was able to handle errors in the data and user input, as well as various scenarios for calculations including the rolling over of an hour or day while calculating the difference in time, or detecting DBMs.

After that experts in the field of data analysis and breathing were approached to comment on the validity of the methodology and approach to breathing classification the program took. The experts approached validated the system in terms of methodology, the way the data was processed and the approach to classify breathing.

Lastly, the program as a whole was tested on participants with prior diaphragmatic breathing experience. The results of the DBM detection test indicate that the program is able to detect DBMs, both scheduled and otherwise if there are no errors introduced while recording data. Care needs to be taken to prevent slippage by fastening the bands in place using a bandage or tape, as tightening them

“How can Respiratory Inductance Plethysmography data be classified and analyzed to be used as feedback towards cultivating habitual diaphragmatic breathing?”

A program was developed to identify periods of diaphragmatic breathing, and to extract various other features regarding diaphragmatic breathing from respiratory data recorded using RIP. Unwanted noise outside the 0.2 - 0.7 Hz frequency range was filtered using a Bandpass filter. Diaphragmatic breathing is classified using a Support Vector Machine to label samples as “Diaphragmatic” or “Chest” as well as “Moving” or “Stationary”, based on accelerometer data. Repetitive breathing is evaluated using breathing frequency, that is extracted using a periodogram and comparing it to the average breathing frequency of previous minutes. Statistics such as a quality factor for the session, the longest DBM, the average DBM for the session etc. are extracted and presented to the user as feedback towards

cultivating habitual diaphragmatic breathing, as these features are of interest. The program was able to identify some DBMs, however errors were introduced due to slippage of the thoracic band. Attention should be paid to the placement and securing of the bands to reduce errors introduced by the slippage of the thoracic or abdominal band.